Apple A11 is 6 core (2+4)

[coercitiv]

If tricore A10X @ 2.4Ghz is faster than dualcore Broawdwell @ 3Ghz in Lightroom and other image processing apps then why isn't the entire industry talking about this?

 

[beginner99]

If tricore A10X @ 2.4Ghz is faster than dualcore Broawdwell @ 3Ghz in Lightroom and other image processing apps then why isn't the entire industry talking about this?

Certain operations could be GPU-accelerated and yeah A10X has a lot faster GPU than intel. So no surprise there.

 

[tamz_msc]

If tricore A10X @ 2.4Ghz is faster than dualcore Broawdwell @ 3Ghz in Lightroom and other image processing apps then why isn't the entire industry talking about this?

Because MacOS!=iOS.

 

[thunng8]

If tricore A10X @ 2.4Ghz is faster than dualcore Broawdwell @ 3Ghz in Lightroom and other image processing apps then why isn't the entire industry talking about this?

There are a couple of video blogs on Lightroom for iOS:

https://www.youtube.com/watch?v=TDvsNRa13CI
https://www.youtube.com/watch?v=xI5tCSMLles

Lightroom is super fast on iOS iPad Pro. It is quite amazing on how fast/smooth it is even editing 42MP RAW files.

Note: I also used to have a Surface Pro 4 and lightroom performance was significantly worse than the macbook pro. It was quite a pain to use. I believe the cause might be that CPU was throttling significantly on the Surface Pro during heavy use.

 

[beginner99]

Note: I also used to have a Surface Pro 4 and lightroom performance was significantly worse than the macbook pro. It was quite a pain to use. I believe the cause might be that CPU was throttling significantly on the Surface Pro during heavy use.

The cause most likely is that Apple paid Adobe to optimize it for the iPad pro specifically because it's 1 device with same hardware. Can't do that with windows laptops.

 

[ksec]

1. The removal of 32bit Hardware is anything but guess right now. We dont have it confirmed ( yet ).

2. Generally Floating Point on ARM, even ARMv8 is awful, or it wasn't ARM's focus. So Intel AVX wins out. But Apple dont implement NEON, VFP or whatever ARM SIMD had off the shelf. Their own implementation are very good, although still not close to AVX ( yet ).

3. And It doesn't have to, the reason why Lightroom and other Apps are running so much better on iPad Pro 10.5 is mostly because of how software and hardware is integrated. Apple Custom GPU, Metal 2, its drivers are all done by Apple and are very much optimized. Compared to on the Mac, Adobe is always late to include any latest macOSX changes such as Metal, mostly because of backward compatibility.

4. A9 was close to Skylake, in some areas and lacking in others. But generally it was close, not ahead of Intel no matter how you spin it to give a fair assessment. A10 is definitely level, as compared to Macbook last year. Some benchmarks and JSC ( Webkit Javascript compiler ) shown that. But A11? It is definately ahead of Skylake, Kabylake, or what ever lake Intel had in mind to milk the bloody PC market.
* However i would bet pure Floating Point with AVX2, Intel still wins.

5. I get that Microsoft is working on ARM64 Windows, I know they promise emulation that Windows on ARM could run most X86 program. But given how Intel would be legally defending it and pressure M$, all these seems so far away.

6. I would love someone take a deep dive into the GPU, and CPU too! But no one on the net are doing these anymore since Anand left.

 

[Eug]

TSMC has been doing very well of late. Given this:

http://www.anandtech.com/show/11337/samsung-and-tsmc-roadmaps-12-nm-8-nm-and-6-nm-added

And this:

http://markets.businessinsider.com/...folio-for-TSMC-7-nm-FinFET-Process-1002358329

Might we expect A12 on TSMC 7 nm next year? Or it too early?

 

[Arachnotronic]

TSMC has been doing very well of late. Given this:

http://www.anandtech.com/show/11337/samsung-and-tsmc-roadmaps-12-nm-8-nm-and-6-nm-added

And this:

http://markets.businessinsider.com/...folio-for-TSMC-7-nm-FinFET-Process-1002358329

Might we expect A12 on TSMC 7 nm next year? Or it too early?

A12 should be TSMC N7, yeah.

 

[StrangerGuy]

The A9 is still a complete monster of a chip. I have never seen any slowdown on iOS10 barring an Safari adblocker bugging out on Amazon.com on my SE.

 

[NTMBK]

I'm quite sad Apple didn't announce an updated iPhone SE. I want to see reasonable sized phones stick around

 

[Eug]

The A9 is still a complete monster of a chip. I have never seen any slowdown on iOS10 barring an Safari adblocker bugging out on Amazon.com on my SE.

People are starting to see some mild slowdowns in iOS 11 with A9 in the iPhone 6s and 6s Plus, but it's still very much OK. I know I'm perfectly happy with the performance (outside of 4K HEVC performance) of my A8X in my iPad Air 2, which is roughly in the ballpark of A9. Also, with my brief testing of my wife's iPhone 6s, it seems fine too. Certainly, my wife hasn't complained at all about iOS 11 performance on that phone.

In contrast, people are complaining about significant slowdowns on A8 in the iPhone 6 and 6 Plus, but the difference there is those are 1 GB devices. My A8X iPad Air 2 not only has an extra core and a faster GPU, it also has 2 GB RAM. The 6s and 6s Plus also have 2 GB RAM.

BTW, my A10 Fusion iPhone 7 Plus feels very responsive with the iOS 11 Gold Master, but that phone not only has a faster CPU, it also has 3 GB RAM (like the 8 Plus and X).

In summary, iOS 11 appears to be targeted for A9 / A9X and up with 2 GB RAM, but works fine on A8X with 2 GB RAM as well (aside from 4K HEVC for the A8X). However, A8 and below with 1 GB RAM are getting problematic.

It's amazing how much power we "need" in our iDevices these days, considering that all of these devices (including even the iPhone 6) are faster in Geekbench than my 2009 13" MacBook Pro, which is still in use. That 2.26 GHz Core 2 Duo P8400 has a multi-core Geekbench 4 score of just 2600, but can still be used with High Sierra.

 

[ksec]

TSMC has been doing very well of late. Given this:

http://www.anandtech.com/show/11337/samsung-and-tsmc-roadmaps-12-nm-8-nm-and-6-nm-added

And this:

http://markets.businessinsider.com/...folio-for-TSMC-7-nm-FinFET-Process-1002358329

Might we expect A12 on TSMC 7 nm next year? Or it too early?

TSMC now pretty much has an Apple iPhone Scheduled roll out. And Apple are willing to have smaller quantity product using leading edge node before iPhone. ( The A10x ) I think this is a partnership that cant be easily replaced. So yes, given 10nm is like the 20nm step before 16nm, next iPhone will highly likely be shipped with 7nm in 2018, and 7nm+ in 2019, possibly 5nm in 2020. ( And that is the reason why I dont understand the rumors of Samsung getting Apple's SoC business, may be for iPad only )

Note. Samsung do have an much ( cost ) optimized 14nm node, which they call 11nm much earlier then TSMC. TSMC 12nm FFC is only scheduled to come out in 2019, while Samsung has it next year. May be some older chip like A9 being improved and Fab on that. The only reason I guess A9 is because A9 is already being manufactured with Samsung 14nm, moving to 11nm makes sense. But the Segment and volume is so small i dont see why it would be cost efficient to do so. Unless Apple have something else in mind.....

Edit: Also notes I think it is in Apple's best interest to keep all their new CPU and GPU design AWAY from Samsung.

 

[StrangerGuy]

The doors to Ax fabbing are forever closed to Samsung, when TSMC is a conflict of interest free party from Apple's perspective barring wafer supply allocations vs NV/QC etc.

As for OLED, trust me when I say Apple is spending tens of billions into LG now to at least get a bargaining chip against Samsung for the next cycle.

 

[Arachnotronic]

The doors to Ax fabbing are forever closed to Samsung, when TSMC is a conflict of interest free party from Apple's perspective barring wafer supply allocations vs NV/QC etc.

As for OLED, trust me when I say Apple is spending tens of billions into LG now to at least get a bargaining chip against Samsung for the next cycle.

Apple gets first dibs to new nodes. Qualcomm has shifted its leading edge business to Samsung ("quid pro quo" to get Snapdragon into Galaxy S), and NVIDIA doesn't bring anywhere near the kind of business (esp. leading edge) to TSMC that Apple does.

TSMC is basically Apple's personal fab at this point.

 

[raghu78]

Apple gets first dibs to new nodes. Qualcomm has shifted its leading edge business to Samsung ("quid pro quo" to get Snapdragon into Galaxy S), and NVIDIA doesn't bring anywhere near the kind of business (esp. leading edge) to TSMC that Apple does.

TSMC is basically Apple's personal fab at this point.

I would say TSMC is Apple's preferred foundry. Both the companies are fueling each other's growth and success .Without TSMC Apple will not be able to deliver the significant yearly improvements we see in the Ax chips and without Apple , TSMC will not have the massive wafer volumes to make leading edge production profitable. . I think this partnership is comparable to the Wintel alliance which dominated the PC industry and made x86 dominant in desktops,notebooks and servers.

 

[Arachnotronic]

I would say TSMC is Apple's preferred foundry. Both the companies are fueling each other's growth and success. Without TSMC Apple will not be able to deliver the significant yearly improvements we see in the Ax chips. I think this partnership is comparable to the Wintel alliance which dominated the PC industry and made x86 dominant in desktops,notebooks and servers.

Yeah. I think part of that is due to TSMC just being a better strategic partner (Samsung is a direct competitor and it's not exactly known for its amazing ethics), and part of that is just that TSMC has a superior logic R&D manufacturing org to Samsung, so going with TSMC means better/denser/more efficient chips than going with Samsung Foundry means.

 

[oak8292]

TSMC also has a larger secondary market for utilization of the foundry equipment after the advanced node players have moved on. Intel moved from a four to a five year depreciation cycle and it shows in how long they are on a node. TSMC is utilizing equipment for two nodes, e.g. 20/16nm and now 10/7nm. Apple will get three years out of a 'dual' node and then there needs to be an adequate secondary market to utilize the remaining years. There is a SemiEngineering article on the surprising number of companies doing tape outs at 7nm where the design costs are starting to get pretty high.

The Apple and Qualcomm wafer volumes have really put new life in the foundries. AMD and Nvidia are beneficiaries of the volumes that mobile has brought.

 

[lopri]

It's not like Apple is known for corporate ethics. Hah. Just ask Imagination. There are a lot more parts than OLED screen made by Samsung in the new iPhones. These corporations seek profits, not score-settling.

 

[Arachnotronic]

It's not like Apple is known for corporate ethics. Hah. Just ask Imagination. There are a lot more parts than OLED screen made by Samsung in the new iPhones. These corporations seek profits, not score-settling.

When I say "corporate ethics" I mean "Samsung can and probably will steal your trade secrets so avoid giving them access to your proprietary IP as much as you possibly can."

Also, how was Apple's handling of the situation with Imgagination "unethical"? Imagination couldn't deliver tech that could keep up with Apple's, which isn't surprising because Apple can invest far more in gfx hardware/software than Imagination could possibly hope to.

 

[raghu78]

Yeah. I think part of that is due to TSMC just being a better strategic partner (Samsung is a direct competitor and it's not exactly known for its amazing ethics), and part of that is just that TSMC has a superior logic R&D manufacturing org to Samsung, so going with TSMC means better/denser/more efficient chips than going with Samsung Foundry means.

True. Apple would be far more comfortable dealing with TSMC than Samsung which directly competes with Apple in the high end smartphone market. I agree that TSMC has better transistor performance, efficiency and yields than Samsung though density is not a measure on which they had the lead at 16F++ . At 10nm TSMC is ahead on density and retains the transistor performance and efficiency lead. For 7nm we need to wait for 2019 to see how TSMC N7+ with EUV and Samsung 7LPP fare on key metrics. It looks like TSMC has regressed on their earlier performance claims wrt N7 and N7+ with EUV against 16FFC.

http://www.eetimes.com/document.asp?doc_id=1332293&page_number=2

"TSMC sketched out what it called a relatively simple process of porting design rules and IP to an N7+ process using EUV that it could put into production in 2019. The process can deliver 20% greater density, 8–10% higher speeds, or 15–20% less power than its current N7 node. Compared to its 16FFC process, N7+ can enable 30% higher speed or 50% less power on an ARM A72 core, said Cliff Hou, vice president of R&D for design technology at TSMC."

"The talks showed that TSMC is working multiple levers to eke out gains, said G. Dan Hutcheson of VLSI Research. “They are getting density, power, and speed improvements — things some people say no longer hold true with semiconductors.

However, some of the results were less impressive than what TSMC estimated back in March, said Mike Demler, senior analyst at the Linley Group. The number of 7-nm tape outs and its performance gains, as well as power savings on 22ULP and performance gains with 12FFC, were all slightly lower than the foundry predicted six months ago, he said."

Six months back

https://community.cadence.com/cadence_blogs_8/b/breakfast-bytes/archive/2017/03/22/tsmc2

"Actually, Cliff Hou, TSMC's VP of Research & Development and Technology Platform, presented in the morning but since his presentation serves in some ways as an introduction to more detailed presentations in the afternoon, I cover it today.

He kicked off by talking about N7 and N7+ readiness. N7 is ready for tapeouts and all tools will be certified by the end of March. Foundation IP (standard cells, memories etc) are ready. Some ecosystem IP (which is TSMC-speak for IP created by other companies such as Cadence) is validated in silicon. N7 (compared to 16FFC) gets either a 33% speed boost or a 58% power reduction (on an ARM core).

N7+, which uses several EUV layers to simplify the process and improve the routing, will get a further 10% performance boost, and the logic will be shrunk a further 15-20%. All design rules remain the same except for the EUV layers, which it seems are more aggressive. So to move from N7 to N7+ will require reimplementation to take advantage of the improved cell libraries, whereas SRAM, analog and I/O will just require re-characterization."

 

[Arachnotronic]

True. Apple would be far more comfortable dealing with TSMC than Samsung which directly competes with Apple in the high end smartphone market. I agree that TSMC has better transistor performance, efficiency and yields than Samsung though density is not a measure on which they had the lead at 16F++ . At 10nm TSMC is ahead on density and retains the transistor performance and efficiency lead. For 7nm we need to wait for 2019 to see how TSMC N7+ with EUV and Samsung 7LPP fare on key metrics. It looks like TSMC has regressed on their earlier performance claims wrt N7 and N7+ with EUV against 16FFC.

http://www.eetimes.com/document.asp?doc_id=1332293&page_number=2

"TSMC sketched out what it called a relatively simple process of porting design rules and IP to an N7+ process using EUV that it could put into production in 2019. The process can deliver 20% greater density, 8–10% higher speeds, or 15–20% less power than its current N7 node. Compared to its 16FFC process, N7+ can enable 30% higher speed or 50% less power on an ARM A72 core, said Cliff Hou, vice president of R&D for design technology at TSMC."

"The talks showed that TSMC is working multiple levers to eke out gains, said G. Dan Hutcheson of VLSI Research. “They are getting density, power, and speed improvements — things some people say no longer hold true with semiconductors.

However, some of the results were less impressive than what TSMC estimated back in March, said Mike Demler, senior analyst at the Linley Group. The number of 7-nm tape outs and its performance gains, as well as power savings on 22ULP and performance gains with 12FFC, were all slightly lower than the foundry predicted six months ago, he said."

Six months back

https://community.cadence.com/cadence_blogs_8/b/breakfast-bytes/archive/2017/03/22/tsmc2

"Actually, Cliff Hou, TSMC's VP of Research & Development and Technology Platform, presented in the morning but since his presentation serves in some ways as an introduction to more detailed presentations in the afternoon, I cover it today.

He kicked off by talking about N7 and N7+ readiness. N7 is ready for tapeouts and all tools will be certified by the end of March. Foundation IP (standard cells, memories etc) are ready. Some ecosystem IP (which is TSMC-speak for IP created by other companies such as Cadence) is validated in silicon. N7 (compared to 16FFC) gets either a 33% speed boost or a 58% power reduction (on an ARM core).

N7+, which uses several EUV layers to simplify the process and improve the routing, will get a further 10% performance boost, and the logic will be shrunk a further 15-20%. All design rules remain the same except for the EUV layers, which it seems are more aggressive. So to move from N7 to N7+ will require reimplementation to take advantage of the improved cell libraries, whereas SRAM, analog and I/O will just require re-characterization."

Thanks for the EETimes link. I'll check it out.

 

[FIVR]

The cause most likely is that Apple paid Adobe to optimize it for the iPad pro specifically because it's 1 device with same hardware. Can't do that with windows laptops.

Microsoft can't afford to optimize specifically for its (3) SP4 skus? The only difference between them is memory size and clock speed. This explanation made me LOL

 

[scannall]

Kind of interesting that Apple sells more CPU's per year than Intel.

 

[beginner99]

Microsoft can't afford to optimize specifically for its (3) SP4 skus? The only difference between them is memory size and clock speed. This explanation made me LOL

I recognized your posts in other threads and really wonder if you are just trolling or believe the non-sense you say.

Obviously Windows version for light room is the same version whether it runs on a surface pro or a dual-socket xeon server with 4 gpus. Yeah the later doesn't make any sense to use for lightroom but it's still the exact same version and it must and will work on both of them. So either stop trolling or stop acting stupid.

 

[ksec]

Kind of interesting that Apple sells more CPU's per year than Intel.

Where did you get this number? Last time I check Apple is close ( in unit ), but hasn't overtaken Intel (yet)*. The last 12 months it is likely Apple has shipped more transistor on CPU then Intel did.

Given how Intel is late with 10nm, PC market is shrinking, DC Market growth is slowing, AMD is gaining some market share, along with an Apple super upgrade, it is highly likely Apple will past Intel in both transistor and CPU unit volume.

And Apple's own GPU's will overtake AMD and Nvidia in unit sales.

Quite Astonishing.

*Apple's Number Excluding Intel Mac and Intel's Numbers include HPC and Data center, Tablet, etc